[Schema] Please rethink model block syntax

model User {
  id    Int    @id
  email String @unique
}

model User {
  id       Int    @id
  email    String @unique
  username String @unique
}

My biggest gripe with the current syntax is how spacing can change every single field in the model. See adding username changes spacing for other fields. This hurts blame for absolutely no reason. I don’t get why:

No pull requests with different spacing schemes.

was such a big deal.

Like gofmt and unlike prettier, there are no options for configurability here. There is exactly one way to format a prisma file.

I am not sure I buy this completely given I don’t know the drawbacks for Schema syntax in v1.34 (latest stable) and what this particular new syntax really adds.

(Early enter. Writing the post)

[matthewmueller]

@sapkra thanks for your thoughts. I agree that for this small datamodel, no formatting is just as readable. Plus it’s simpler.

While evaluating our decision, we’ve been looking at large datamodels that you’d find in a real app:

I think the column alignment improves readability quite a bit.

[auderer]

I notice a pattern here that I’ve noticed in many other areas of software development: separation of concerns.

The formatted version is sliced vertically.

The non-formatted version is sliced horizontally.

Which one is more useful when actually working in a large dataset? I would argue that horizontal organization is far more important. When glancing at a screenshot of code, columns look far more aesthetically pleasing; but that’s not how we actually read our data when editing, adding, or debugging.

We don’t vertically scan columns of types or tags. We scan field names, then look ahead horizontally to see which types and tags apply to that field. This is markedly more difficult in a column format, especially when there’s a large gap between the longest and shortest field names.

How does this relate to separation of concerns? It reminds me of organizing HTML/CSS/JS by file-type instead of by component. Each line in the model is like a component - it has a name, type, and decorators. These three values should be closely coupled together and independently organized; not spread out based on the position of values in other rows.

I definitely see how column formatting appears more readable, but in practice it is not. If I have to drag my cursor across the line or hold a ruler up to my screen just to see which type/decorators apply to a field, it’s a bad design. Not to mention the terrible diffs and merge conflicts that this causes.